Recipe generation of under fill process based on improved kernel regression and particle swarm optimization

Abstract

The under fill process is a process that fills the gap between a chipset and a substrate using an epoxy material. The output of this process is a length of tongue that has to be controlled so it avoid touching the keep out zone. A recipe generation of the input parameters in the under fill process will help the length of tongue generated from touching the keep out zone. This project proposes a predictive modeling algorithm called Improved Kernel Regression and Particle Swarm Optimization in order to find the six input parameters needed in the under fill process. Even though only few samples of the under fill data sets are used in the simulation experiment, the proposed approach is able to provide a recipe generation of the six input parameters.